Impact of Decipher on use of post‐operative radiotherapy: Individual patient analysis of two prospective registries

Abstract Objective To assess the association between Genomic Classifier (GC)‐risk group and post‐radical prostatectomy treatment in clinical practice. Methods Two prospective observational cohorts of men with prostate cancer (PCa) who underwent RP in two referral centers and had GC testing post‐prostatectomy between 2013 and 2018 were included. The primary endpoint of the study was to assess the association between GC‐risk group and time to secondary therapy. Univariable (UVA) and multivariable (MVA) Cox proportional hazards models were constructed to assess the association between GC‐risk group and time to receipt of secondary therapy after RP, where secondary therapy is defined as receiving either RT or ADT after RP. Results A total of 398 patients are included in the analysis. Patients with high‐GC risk were more likely to receive any secondary therapy (OR: 6.84) compared to patients with low/intermediate‐GC risk. The proportion of high‐GC risk patients receiving RT at 2 years post‐RP was 31.5%, compared to only 6.3% among the low/intermediate‐GC risk patients. Conclusion This study demonstrates that physicians in routine practice used GC to identify high risk patients who might benefit the most from secondary treatment. As such, GC score was independent predictor of receipt of secondary treatment.

30% of the time. 9 The arguments against aRT administration are: the high number needed to treat, increased treatment-related toxicity, and decreased quality-adjusted life expectancy. 10 Therefore, a more sagacious approach of surveillance and administering early salvage radiation (sRT), when prostate specific antigen (PSA) is detectable but ideally less than 0.5 is more embraced among the urological community. 11 The evidence to support the use of early sRT is based on three recently presented randomized trials, RADICALS (ISRCTN40814031), GETUG-AFU 17 (NCT00667069) and RAVES (NCT00860652), which showed that early sRT is not inferior to aRT. [12][13][14] The use of Genomic Classifier (GC) data in the post-radical prostatectomy have been shown to predict the risk of metastasis and prostate cancer-specific mortality, as well as improve the accuracy of risk stratification of high-risk prostate cancer. 15 Hence, it identifies who might benefit from aRT, which might impact patient as well as physician decision-making regarding additional treatment. 16 In this study, we aim to assess the association between GC risk group and post-radical prostatectomy treatment in real world practice via utilization of two contemporary prospective cohorts of men treated with RP.

| Study cohorts
This study adheres to the REMARK criteria for the evaluation of prognostic biomarkers. 17 This study included two prospective observational cohorts of men diagnosed with prostate cancer (PCa) that were treated with RP and had available post-RP GC results. GC testing was offered to patients with adverse pathological features on the final pathology (eg, positive surgical margin, presence of extraprostatic extension, seminal vesicle invasion, or PSA persistence). 18 The first cohort is the University of Pennsylvania (UPenn) cohort, which is part of a prospectively collected Institutional Review Board (IRB) approved comprehensive database for Robot-Assisted Radical Prostatectomy (RARP) maintained by the corresponding author. The second is the Thomas Jefferson University (TJU) cohort, which is an IRB approved prospective cohort that implemented GC-based treatment recommendations. 19 The UPenn cohort consisted of 352 patients who underwent RARP between 2013 and 2018, while the TJU cohort consisted of 135 patients treated by RP between 2014 and 2016. In the UPenn cohort, the decision and the timing to administer adjuvant/early salvage radiation therapy and androgen deprivation therapy (ADT) were based on patient co-morbidities and life expectancy, patients' treatment expectations, PSA kinetics, and consensus of the prostate cancer multidisciplinary team. In the TJU cohort, the decision and timing to administer aRT, with or without ADT, was based on tumor board recommendations pertaining to the genomic risk score. 17 Patients were excluded from analysis if they developed biochemical recurrence before the GC test, had lymph node invasion, PSA persistence, missing follow-up, or pathologic information, or if they received any neoadjuvant therapy or non-concurrent adjuvant ADT, defined as receiving ADT more than 6 months before RT. ( Figure S1).

| GC and CAPRA-S
Decipher test results were collected prospectively in the registry, but in order to compare GC continuous score across versions of the test, GC scores were retrieved directly from the Decipher Genomics Resource Information Database (GRID). GC scores were calculated based on the predefined 22-marker Decipher classifier. 20 The GC score is a continuous score between 0 and 1, with the lowest scores indicating a lower risk of metastasis. GC score categorical low-, intermediate-, or high-risk were based on pre-specified cutpoints of <0.45, 0.45-0.6, or >0.6, respectively. 15 The Cancer of the Prostate Risk Assessment Post-Surgical (CAPRA-S) scores were calculated using pre-operative PSA, pathological Gleason score, surgical margin status, extraprostatic extension, seminal vesicle invasion, and lymph node invasion. 21 The CAPRA-S score is an ordinal score between 0 and 12, with the lowest scores indicating a lower risk of PCa recurrence post-RP. Patients were categorized as low-, intermediate-, or high-risk based on CAPRA-S scores of ≤2, 3-5, or ≥6, respectively.

| Endpoints
The primary endpoint of the study was a priori chosen as time to receipt of secondary therapy after RP, where the objective was to assess the association between GC-risk group and time to secondary therapy. Secondary therapy is defined as receiving either RT or ADT after RP and time to secondary therapy is calculated from time of RP until event (the first occurrence of RT or ADT) or last follow-up. A secondary endpoint was time to biochemical failure (BF) or receipt of salvage ADT (sADT). Biochemical failure is defined as PSA ≥ 0.2 ng/ mL post-RP secondary therapy. Salvage ADT is defined as receiving ADT after biochemical recurrence (BCR) or (at least 6 months) after RT, whereas ADT administered within 6 months of RT was considered concurrent therapy. Time to biochemical failure or receipt of sADT is calculated from time of RP until event (the first occurrence of BF or sADT) or last follow-up. aRT was defined by the initiation of radiation prior to BCR, while sRT was defined by the initiation of radiation after BCR.

| Statistical analysis
Descriptive statistics of the two cohorts combined are reported by medians and interquartile-ranges (IQR) or frequencies and proportions, as appropriate. Distribution of radiation therapy usage across CAPRA-S risk groups were compared using Fisher's exact test within SHAHAIT eT Al.

| 268
GC risk groups. 22 Kaplan-Meier cumulative incidence curves of the receipt of secondary therapy risk were constructed and compared using the log-rank test. 23 Univariable (UVA) and multivariable (MVA) Cox proportional hazards models, stratified by institution, were used to evaluate the prognostic value of GC scores and individual clinicopathologic risk factors in predicting the risk of receiving secondary ing in a total of 398 patients for analysis ( Figure S1). risk CAPRA-S were found to have low-GC scores (Figure 1).  (Table S2) Overall, secondary therapy was administered at different rates when stratified by GC risk (Figure 2, P < .001), where high-GC risk F I G U R E 2 Cumulative incidence of the receipt of secondary therapy stratified by GC risk, compared using the log-rank test. ADT, androgen deprivation therapy, RP, radical prostatectomy, RT, radiation therapy SHAHAIT eT Al.  (Table S3).

| Timing of RT for different GC-risk groups and biochemical failure or receipt of salvage ADT
Overall, patients developed BF or received sADT at different rates when stratified by GC risk (Figure S3

| D ISCUSS I ON
In this study, we found that patients with high-GC risk were 6.84 times more likely to receive any secondary therapy compared to patients with low/intermediate-GC risk and that GC score was an independent predictor of receiving secondary therapy. This shows that more physicians use GC score not only for risk stratification and prognostication but to Training registry cohort. 27 However, its applicability to the daily practice has been debated. In this report, we confirm the versatility Interestingly, we found that aRT mitigated the effect of GC on the risk of developing BF or receiving sADT. In this study we sought to use BF as a surrogate endpoint for analysis, as it is highly associated with future risk of metastasis. 28 The ARTISTIC meta-analysis of three randomized trials, RADICALS (ISRCTN40814031), GETUG-AFU 17 (NCT00667069), and RAVES (NCT00860652), showed that early sRT is not inferior to aRT. 29 It is noteworthy that these trials failed to account for the molecular heterogeneity of prostate cancer. Additionally, the number of patients with multiple adverse pathological features was low, which affected the applicability of the results from these trials on a very high-risk population with multiple adverse pathological features and high-GC risk. Indeed, these individuals are most likely to benefit from aRT, rather than sRT, as previously demonstrated by Dalela et al. 30 Our study has several limitations that warrant discussion. First, the short median follow-up of 2 years necessitated analysis of post-operative treatment failure, a surrogate endpoint, rather than a proven surrogate endpoint for survival, such as metastasis. Moreover, the number of patients who received sRT was small and the fact that this was an observational registry (ie, treatment was not randomly assigned, but based on baseline clinical and genomic risks) did not allow us to do a formal comparison with aRT. Finally, GC testing at UPenn was offered to all patients with pT3 disease, and pT2 with positive surgical margins; however, some patients declined the test due to lack of insurance coverage.